Re: drivers/md/dm-cache-metadata.c:1705:9: warning: 'strncpy' specified bound 16 equals destination size

[Justin Stitt]

On Sat, Sep 30, 2023 at 2:49 AM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>
> On Thu, Sep 28, 2023 at 11:36:46AM +0200, Miguel Ojeda wrote:
> > On Thu, Sep 28, 2023 at 12:24 AM kernel test robot <lkp@xxxxxxxxx> wrote:
> > >
> > >    In function 'write_hints',
> > >        inlined from 'dm_cache_write_hints' at drivers/md/dm-cache-metadata.c:1729:6:
> > > >> drivers/md/dm-cache-metadata.c:1705:9: warning: 'strncpy' specified bound 16 equals destination size [-Wstringop-truncation]
> > >     1705 |         strncpy(cmd->policy_name, policy_name, sizeof(cmd->policy_name));
> > >          |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >
> > This looks fine given the check above it. In any case, since it
> > appears the code wants to terminate the string, this should use
> > `strscpy*()`, right?
>
> I agree. I can't tell if _pad is needed though. I think struct
> dm_cache_metadata is only used internally? And I see at least the
> initial allocation is zeroed:
>
>         cmd = kzalloc(sizeof(*cmd), GFP_KERNEL);
>
> Regardless, for background,
>
> struct has:
>         char policy_name[CACHE_POLICY_NAME_SIZE];
>
> code does:
>         const char *policy_name = dm_cache_policy_get_name(policy);
>
>         if (!policy_name[0] ||
>             (strlen(policy_name) > sizeof(cmd->policy_name) - 1))
>                 return -EINVAL;
>
>         strncpy(cmd->policy_name, policy_name, sizeof(cmd->policy_name));
>
>
> Justin, can you add this to your queue (if it's not already there)?

I sent a patch out earlier this week [1].

The line # on the diff v.s the warning is off though. I suspect I may
need a rebase if it doesn't apply clean.

>
> -Kees
>
> --
> Kees Cook

[1]: https://lore.kernel.org/all/20230925-strncpy-drivers-md-dm-cache-metadata-c-v1-1-4b75c7db0cfe@xxxxxxxxxx/